How to make a success out of SPC
Most people must be convinced on their home turf that statistical process control (SPC) is worth it. Skeptics might say, “Isn’t SPC another of those Japanese business practices, like lifetime employment, morning calisthenics, company songs, and quality circles? Must we try all of them?”
Some attitudes can be well-justified and dead wrong at the same time. It’s true, over the past five years we’ve sent plenty of people to Japan to study their business practices. After a few weeks, sometimes months, they come back with all sorts of theories. It’s enough to make you a little wary. The problem with this viewpoint is that it just isn’t true about SPC. First of all, its pioneers are a pair of Americans, W.A. Shewhart and W.E. Deming. Although it seemed discredited in the 1950s and 1960s, when American industry had little competition for its products, SPC is the management tool to achieve and maintain product quality. William Conway, former chairman of Nashua Corporation of Massachusetts called Dr. Deming’s methods the third wave of the industrial revolution. The first wave was the mechanization of English textile plants in the 18th century. The second wave was the introduction of mass production by Henry Ford.
Strong testimonies notwithstanding, there are some definite problems with SPC. They have to do with that dark, fearsome beast called implementation. When you read about Dr. Deming’s management philosophies, and study SPC as it is currently practiced in Japan, you find out what a rigorous practice it can be. Some of the conditions deemed necessary for SPC to work are
- Educating management on the philosophy of SPC.
- Training management and hourly workers to use simple statistics and control charts. (Sometimes basic mathematics must be taught first.)
- Retooling fixtures and inspection systems to measure variability. (Reducing or eliminating final inspection is an eventual goal.)
- Dedicating workers, process engineers, and management to the task of process improvement based on statistical techniques. (This can involve restructuring jobs, creating work standards, and even developing wage contingencies.)
- Accepting the results — which can affect product deliveries, inventory, tooling, machinery purchase, and the whole manufacturing process.
As they say, strong medicine sometimes kills the patient. What perhaps ought to be whispered while getting started is: “You can start small.” Most of the examples mentioned here involve American companies that have really just started. They didn’t send all their personnel to statistics boot camp. They assigned responsibilities to a small group, usually the quality or process engineering group. For fear of choking, they didn’t swallow the whole retooling argument either. What they did do was pinpoint one or two processes that could stand improvement, and they gave SPC a try.
Edgewood Tool and Manufacturing Co. of Taylor, Michigan, is one example. The company had a problem with misformed parts on hood hinges for Ford light trucks. The problem was traced back to the blanking stage. One critical dimension — the distance from the edge of a pierced hole to the edge of the part — was monitored using a control chart. The company found that parts variation increased whenever the operator loaded a new coil onto the machine. The solution was an inexpensive gaging block that made loading and positioning the coil a more precise operation.
The significance of Edgewood Tool’s use of statistical process control is not that the company was able to solve a problem and decrease scrap, rework, and inspection, but that it was done by monitoring only one characteristic. Only a few trained personnel were needed to achieve results.
A bumper supplier for Ford Motor Company had a similar experience. The supplier used SPC to monitor a plating bath. By determining the capability of the plating bath, and working on one characteristic — bath temperature — a 25% energy savings was realized on that part of the process. Once again, a limited application paid off both in real terms and in an increased awareness of the power of SPC.
Ford, in its effort to encourage its suppliers to implement SPC, offers these suggestions:
- Start with a pilot program.
- Select just a few characteristics.
- Train the involved group in statistics.
- Adopt gaging for variability.
- Document your results.
- Create management awareness
Ford also suggests purchasing a data collector/quality control computer. A computer will offer considerable help in collecting data and making statistical sense out of it — areas in which newcomers to SPC can have quite a few problems. But more on the hardware later. First, let’s expand Ford’s suggested methodology.
Starting a Pilot Program
The chief criteria for a pilot program should be its visibility and its potential for success. There is an advantage in selecting a new operation for these reasons. A new operation may also have newer machinery, and the workers may be more inclined to try new methods. If an older process is used, select one that needs improvement, and where improvement can be measured. As many variables as possible should be isolated, so that the results of your work will be dramatic.
Selecting a Few Characteristics
There is a tendency for newcomers to try to monitor as many characteristics as possible. ITT Hancock, a supplier of front door hinges for Ford, had this problem. ITT’s quality control manager and supervisor attended a five-day seminar on statistics conducted by the American Supplier Institute (formerly Ford Supplier Institute), and came away determined to implement a program. The initial approach was to try to control 28 characteristics that contributed to a hinge torque problem. The company found that it could make little headway. After three months, the number of characteristics was cut to 12 and then to 5 for ongoing control.
One of ITT Hancock’s conclusions was, “ . . . an SPC program will generate the earliest results and enthusiastic local support when only a few characteristics are selected for a first application.” Tangible gains included a 10% reduction of machine downtime, a 15% reduction of labor for rework, a 10% reduction of scrap, and improved relations with Ford. But the company also learned a bit about implementing SPC by the experiences of its own operation.
Training in Statistics
Training should include some of the prime decision makers involved in your pilot program. That would normally include the QC manager, production manager and supervisors, and plant manager. Since, according to Dr. Deming, management is responsible for 85% of quality problems, upper management training is needed to provide the rationale for change when change is required.
Those directly responsible for the pilot program, including the production engineer, QC auditor, and machine operators, should also receive some training. The training should include how to make and keep control charts and how to correctly interpret results. This would help eliminate the common pitfalls of making corrective adjustments too soon in a process.
If you are an automotive supplier, excellent training can be obtained through the American Supplier Institute seminars. Unfortunately, most other industries have not yet developed extensive training resources. Hiring a statistical consultant to help with the program is one solution.
Adopting Gaging for Variability
Part of the selection process for the pilot program must involve choosing characteristics that can be measured easily, and to which you can apply statistical techniques. The aim of SPC is to reduce variability, regardless of engineering specifications. This means that gaging and fixturing must be capable of measuring any significant deviation from the nominal. GO/NO-GO gages and other pass-fail types of fixtures will not do.
It is in the accuracy of measurement devices that the power of statistics to identify problems lives or dies. It is here where you should take advantage of some of the latest measurement technology. Since it is necessary to collect data systematically and render it meaningful with the least amount of error, the best system would be one that ties measuring devices directly to data collection equipment. Gages with readouts still require you to copy down readings. They invite transcription errors and considerable intermediate paperwork. The ability to handle the data collection dictates to a large extent the number of characteristics you can control. You can imagine how bogged down in paperwork an ambitious SPC program can get.
Documenting Your Results
The primary document used is the control chart. It takes some training to read one properly; however, once understood, control charts are used to get the process in control. Once in control, the process can use histograms and capability studies to predict whether design tolerances can be met.
Histograms and capability studies are documents that managers will find valuable. With them you can estimate the percentage of defective parts. How much scrap or rework is expected will determine your inspection load. You can also use histograms to minimize the cost of production. Presuming that scrap is more costly than rework, minimize production cost by shifting the mean of your parts distribution away from the scrap end, and accepting just a bit more rework.
Creating Management Awareness
This suggestion is the most important. The selling aspect of SPC pilot programs cannot be overemphasized. You must create management awareness by making a success story. Most companies have plenty of experience with the problemridden periods of process startup. One strong point right off the bat is that SPC is a morale-builder. Workers feel that they gain a controlling influence over the process. Most companies using SPC find that the workers become the best advocates of it. That’s why publicity about the program should flow upward to management and to other areas of your company that could use SPC.
A well-documented SPC program provides the kind of statistics and charts that managers can understand and translate into dollars. The Pontiac Division of General Motors, for example, boasts that implementation of SPC in one plant cut the cost of engine production by 30% in 18 months.
Choosing a System
There are a number of computer-aided SPC systems on the market today. Most require a desktop computer and software to perform analysis. They depend on data to be manually input, which usually produces results slower than you would want. Automated data collection systems solve that problem.
Automated data collection systems quickly collect and process data, providing control charts, capability studies, and histograms simply by connecting to a computer or printer. No other equipment is required to generate graphs and reports. One QC auditor can perform both the data collecting and report generation.
An automated system can manage your data collection and calculate the statistics for you, leaving you free to concentrate on analysis and problem-solving. You can build a bit more potential for success into your pilot program. An automated data collection system also provides a systematic approach to data gathering that you can literally carry over to other manufacturing processes.
To repeat, American companies so far have achieved quite a bit of success using SPC. This is in spite of most of it being in its infancy — limited to one or two processes and a small group of people. In many ways, success is assured by this approach, because any negative ramifications will never be more than what is involved with the trials of improving a single process.
Using a pilot program makes the learning process easier, and it cuts through the hierarchy of job responsibilities by making part production the main goal, and process control a unified endeavor. By using current technology right from the start, and by promoting the success to others in your organization, a pilot program can spearhead the greater usage of SPC and the revitalization of quality control in your company.